Jetson Nano Plate recognition CUDA problem

[musa.karagoz]

Hi everyone,

I’ve been working on this issue for a while. I enabled CUDA. It works on ubuntu 18.04 with version 10.2. You can see the code I wrote below. CUDA not available error persists when easyocr is active.
Whenever I deactivate the codes containing easyocr by leaving # at the beginning, the camera opens and I can easily take images without decreasing the FPS. Whenever I activate easyocr the FPS is around 1-2. I can get CUDA 10.2 output with nvcc --version check. Am I not supposed to use easyocr here? Should I continue with a different library? How can I get license plate reading without FPS drop with CUDA support?
can you please help me with this?
In addition, I install the libraries by installing pycharm on jetson nano and using the pip install command.

import cv2
import datetime
import imutils
import numpy as np
from unittest import result
import easyocr

cap = cv2.VideoCapture(0)

def rec(img):
#Filtreleme islemi
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
gray = cv2.bilateralFilter(gray, 11, 17, 17)
edged = cv2.Canny(gray, 30, 200)
try:
cnts = cv2.findContours(edged.copy(), cv2.RETR_TREE,cv2.CHAIN_APPROX_SIMPLE)
cnts = imutils.grab_contours(cnts)
cnts = sorted(cnts, key=cv2.contourArea, reverse=True)[:10]
screenCnt = None
for c in cnts:
peri = cv2.arcLength(c, True)
approx = cv2.approxPolyDP(c, 0.017 * peri, True)
if len(approx) == 4:
screenCnt = approx
break
if screenCnt is None:
detected = 0
else:
detected = 1

    if detected == 1:
        cv2.drawContours(img, [screenCnt], -1, (0, 255, 0), 3)
    #Maskeleme islemi
    mask = np.zeros(gray.shape, np.uint8)
    new_image = cv2.drawContours(mask, [screenCnt], 0, 255, -1,)
    new_image = cv2.bitwise_and(img, img, mask=mask)
    #Kırma islemi
    (x, y) = np.where(mask == 255)
    (topx, topy) = (np.min(x), np.min(y))
    (bottomx, bottomy) = (np.max(x), np.max(y))
    Cropped = gray[topx:bottomx + 1, topy:bottomy + 1]
    #EasyOCR islemi
    reader = easyocr.Reader(['en'])
    text = reader.readtext(Cropped)
    text = text[0][1]
    #print(text)

    '''f = open("plate.txt", "r")
    result = f.read().find(text)
    f.close()
    if result == -1:
        time = datetime.datetime.now()
        f = open("plate.txt", "a")
        f.write(text + " ")
        f.write(str(time) + "\n")
        f.close'''

    text = text.replace(" ", "")
    text = text.upper()
    sonuc = 0
    if(len(text)>=7):
        if(int(text[0]) and int(text[1])):
            if(str(text[4])):
                if(int(text[5:])):
                    sonuc=1
            else:
                if(int(text[4:])):
                    sonuc=1
    if(sonuc==1):
        f = open("plate.txt", "r")
        result = f.read().find(text)
        f.close()
        if result == -1:
            time = datetime.datetime.now()
            f = open("plate.txt", "a")
            f.write(text + " ")
            f.write(str(time) + "\n")
            f.close
        print(text)
except Exception:
    pass
return img
pass

while True:
ret, frame = cap.read()
img = rec(frame)
cv2.imshow(‘Screen’, img)
if cv2.waitKey(20) & 0xFF == ord(‘q’):
break

cap.release()
cv2.destroyAllWindows()

[AastaLLL]

Hi,

Do you meet an error or is performance drop when enabling the EasyOCR?
EasyOCR is a DNN-based classifier so it’s expected to take some resources and time.

Please check the below link for the info of EasyOCR package with CUDA support:

installing on jetson nano with cuda support

Thanks.